Re: SIGSEGV signal on Darwin OS X (Intel) when using version matching (nmap 4.01)

["Kelly M" <bangpath () sent as>]

On 3/7/06, Fyodor <fyodor () insecure org> wrote:
> On Mon, Mar 06, 2006 at 08:26:41PM -0800, Kurt Grutzmacher wrote:
> > Easy fix is to compile libpcre outside of nmap with utf8 support and
> > then build nmap against that library instead. Not entirely sure why
> > utf8 is required, just that it's the only way to get pcre to
> > function. I used darwinport's pcre and it works great.
>
> For you it works when you use a recompiled (w/UTF-8 support) libpcre.
> For Kelly it works when he gives the debugging (-g) flag to the
> compiler.  In both cases, I think memory is just being rearranged such
> that the problem doesn't manifest itself by crashing in the same way.

In a completely unrelated project (to start learning ruby on rails), I
built libpcre using CFLAGS=-01 (which seems to be required for Intel
Macs) . Later that same night I compiled Nmap again with the -g
debugging flag, so I could post the stack trace here. But Nmap
wouldn't crash anymore. :) Confused about this, I removed the flag and
compiled Nmap again... still didn't crash. So I was miffed. But also
happy that things were working fine. I've been doing lots of scanning
to get that trace to you, but still no dice. :) So that must indeed be
a workaround for now.

> But the underlying problem is likely still there and it could start
> crashing with different flags or with the next release.  So while
> these workarounds may be useful for the moment in allowing you to use
> version detection, we really need to track down the problem and find a
> real solution.  Do we have any volunteers (with an OS X Intel box) to
> give it a shot?

Sure I can help out. Just let me know what I can do.

Kelly


_______________________________________________
Sent through the nmap-dev mailing list
http://cgi.insecure.org/mailman/listinfo/nmap-dev